Hello!
I am trying to automate classification of discharge lamps' spectra. I have a table of possible spectra and for each lamp a list of refraction angles for different spectral lines, however, some lines can be missing due to their low intensity or because they are too far in the violet or in the red zone. The goal is to automatically choose the best-fitting column of the table for each spectrum given. The problem is that I can't represent spectra as points in n-dim Euclidian space, because n can be different and, as I mentioned, some lines can be omitted, which may lead to dramatical changes in distance between the points. By the same reason I can't utilise linear correlation coefficient. So, could you please tell me if there is a common way to evaluate how similar are two real-valued sequences of different lengths? Thank you in advance.
P.S. My English can sometimes be as good as Mark Twain's Deutsch, I apologize for that. Please, let me know if you find any mistakes